Biological data retrieval from web resources often necessitates multi-step access to multiple information sources. User-designed scenarios are exploited by a generic application (Xcollect) that allows users to execute them and to store the collected data in a document. Multiple answers are obtained at given steps of the scenario when several resources are queried with same purpose. We address the problem of managing such multiple answers when retrieving functional annotations of genes. Relevant quality metadata for sources and source entries have been listed in view of sorting the answers. Work in progress deals with semantic integration based on domain ontologies.